BulletGen: Improving 4D Reconstruction with Bullet-Time Generation

arXiv:2506.18601v2 Announce Type: replace-cross Abstract: Transforming casually captured, monocular videos into fully immersive dynamic experiences is a highly ill-posed task, and comes with significant challenges, e.g., reconstructing unseen regions, and dealing with the ambiguity in monocular depth estimation. In this work we introduce BulletGen, an approach that takes advantage of generative models to correct errors and complete missing information in a Gaussian-based dynamic scene representa